BREAST

Radiological review of prior screening mammogramsof screen-detected breast cancer

Tone Hovda1,2 & Kaitlyn Tsuruda3,4 & Solveig Roth Hoff5,6 & Kristine Kleivi Sahlberg7,8& Solveig Hofvind3,9

Received: 23 March 2020 /Revised: 28 May 2020 /Accepted: 31 July 2020# The Author(s) 2020

AbstractObjective To perform a radiological review of mammograms from prior screening and diagnosis of screen-detected breast cancerin BreastScreen Norway, a population-based screening program.Methods We performed a consensus-based informed review of mammograms from prior screening and diagnosis for screen-detected breast cancers. Mammographic density and findings on screening and diagnostic mammograms were classified accord-ing to the Breast Imaging-Reporting and Data System®. Cases were classified based on visible findings on prior screeningmammograms as true (no findings), missed (obvious findings), minimal signs (minor/non-specific findings), or occult (nofindings at diagnosis). Histopathologic tumor characteristics were extracted from the Cancer Registry of Norway. TheBonferroni correction was used to adjust for multiple testing; p < 0.001 was considered statistically significant.Results The study included mammograms for 1225 women with screen-detected breast cancer. Mean age was 62 years ± 5 (SD);46% (567/1225) were classified as true, 22% (266/1225) as missed, and 32% (392/1225) as minimal signs. No difference inmammographic density was observed between the classification categories. At diagnosis, 59% (336/567) of true and 70% (185/266) of missed cancers were classified as masses (p = 0.004). The percentage of histological grade 3 cancers was higher for true(30% (138/469)) than for missed (14% (33/234)) cancers (p < 0.001). Estrogen receptor positivity was observed in 86% (387/469) of true and 95% (215/234) of missed (p < 0.001) cancers.Conclusions We classified 22% of the screen-detected cancers as missed based on a review of prior screening mammogramswithdiagnostic images available. One main goal of the study was quality improvement of radiologists’ performance and the program.Visible findings on prior screening mammograms were not necessarily indicative of screening failure.Key Points• After a consensus-based informed review, 46% of screen-detected breast cancers were classified as true, 22% as missed, and32% as minimal signs.

• Less favorable prognostic and predictive tumor characteristics were observed in true screen-detected breast cancer comparedwith missed.

• The most frequent mammographic finding for all classification categories at the time of diagnosis was mass, while the most frequentmammographic finding on prior screening mammograms was a mass for missed cancers and asymmetry for minimal signs.

Keywords Mass screening . Breast neoplasm . Digital mammography .Mammography . Female

* Solveig Hofvindsolveig.hofvind@kreftregisteret.no

1 Department of Radiology, Vestre Viken Hospital Trust, PO Box 800,3004 Drammen, Norway

2 Institute of Clinical Medicine, University of Oslo, PO Box 1171,Blindern, 0318 Oslo, Norway

3 Section for Breast Cancer Screening, Cancer Registry of Norway, POBox 5313, Majorstuen, 0304 Oslo, Norway

4 Oslo Centre for Biostatistics and Epidemiology, Department ofBiostatistics, Institute of Basic Medical Sciences, University of Oslo,PO Box 1122, Blindern, 0317 Oslo, Norway

5 Department of Radiology, Ålesund Hospital, Møre og RomsdalHospital Trust, Åsehaugen 5, 6017 Ålesund, Norway

6 Department of Circulation and Medical Imaging, Faculty ofMedicine and Health Sciences, NTNU, Trondheim, Norway

7 Department of Research and Innovation, Vestre Viken HospitalTrust, PO Box 800, 3004 Drammen, Norway

8 Department of Cancer Genetics, Institute for Cancer Research, OsloUniversity Hospital Trust, PO Box 4950, 0424 Oslo, Norway

9 Faculty of Health Science, Oslo Metropolitan University, PO Box 4,St. Olavs Plass, 0130 Oslo, Norway

https://doi.org/10.1007/s00330-020-07130-y

/ Published online: 1 October 2020

European Radiology (2021) 31:2568–2579

AbbreviationsBI-RADS Breast Imaging-Reporting and Data SystemCI Confidence intervalCRN Cancer Registry of NorwayDCIS Ductal carcinoma in situDM Digital mammographyER Estrogen receptorHER-2 Human epidermal growth factor receptor 2ILC Invasive lobular carcinomaIQR Interquartile rangeNST No special typePACS Picture archiving and communication systemPR Progesterone receptorSD Standard deviation

Introduction

Breast cancer is diagnosed among screening participants asscreen-detected breast cancer or interval breast cancer (breastcancer diagnosed between two scheduled screening roundsafter a negative screening episode). When obvious mammo-graphic findings corresponding to the location of the tumor arevisible on prior screening mammograms, which in retrospectshould have resulted in a recall, the cancer may be defined asmissed. In studies, up to 50% of interval and screen-detectedbreast cancers can present visible findings on prior screeningmammograms, ranging from minor benign-looking findingsto obviously missed cancers [1–8]. Breast cancer can bemissed at screening due to misperception—the lesion is notperceived by the radiologist—ormisinterpretation—the lesionis detected by the radiologist, but not considered suspiciousenough to warrant a recall, either by the reading radiologist orat a consensus meeting. Further, unsatisfactory image quality,positioning, or inadequate assessment at recall may cause acancer to be missed [9, 10]. Mammographic density may alsoimpact the rates of missed breast cancer due to the maskingeffect of dense breast tissue and overlapping structures [11].

Radiologic reviewsmay be useful for quality assurance andquality improvement of both the program and the radiologists.European guidelines [12] and BreastScreen Norway qualitymanual [13] recommend continuous surveillance and regularreview of screening mammograms performed prior to diagno-sis of interval breast cancer. Further, in the National HealthService Breast Screening Programme (UK), radiologists areobliged to audit mammograms of interval breast cancer [14,15]. Use of audits is discussed also in other countries, and weexpect this topic to receive more attention in the future [16].However, the information available as well as the number ofreviewers affect the results of the review [3, 4]. Further,whether a commitment to inform the women will affect theresults of a review or an audit is debatable.

Several review studies of interval breast cancer have beenperformed [8, 17–20]. Larger review studies of screen-detected breast cancer, particularly including digital mam-mography (DM), are, to our knowledge, sparse. Screen-detected breast cancer with no visible findings on prior screen-ing mammograms, defined as true cases, may grow faster thanmissed breast cancer. Thus, different histopathological char-acteristics and different mammographic findings are anticipat-ed for true versus missed screen-detected breast cancers.

We conducted a nationwide consensus-based, informed re-view within BreastScreen Norway. The study included priorscreening mammograms and mammograms available at diag-nosis from women diagnosed with screen-detected breast can-cer. The overall aim of the study was quality improvement forradiologists’ performance and the program as such. The ob-jectives were to investigate the proportions of true and missedscreen-detected breast cancers and to explore whether mam-mographic findings, density, or histopathological characteris-tics differed between the two groups. We hypothesized thatthese three aspects differed between true and missed screen-detected breast cancers.

Materials and methods

The study was approved by the data protection official at theCancer Registry of Norway (CRN) (PVO approval number:2016/4696), and the local breast centers agreed to the study.The Cancer Registry Regulation waived the requirement toobtain written informed consent for use of screening data forquality assurance and research [21].We received de-identifieddata for analyses from the CRN.

BreastScreen Norway offers women aged 50–69 biennialscreening with two-view standard DM. The screening examstake place at 27 stationary or mobile units. Screen reading isperformed at 16 breast centers and includes independent dou-ble reading by breast radiologists. The radiologists score eachbreast on a 5-point scale; 1 indicates negative findings, and 5indicates a high suspicion for malignancy. Exams scored ≥ 2by either radiologist are discussed in a consensus meeting todecide whether to recall the woman [22]. The median annualreading volume for radiologists during 1996–2016 was 4492exams; 46% of the radiologists had over 10 years of screenreading experience [23].

Materials and review logistics

CRN randomly extracted 85 screen-detected breast cancercases from each of the 16 breast centers. All examinationswere performed with DM during 2006–2015, and all womenhad a screening exam with DM 2 years previously (priorscreen). We aimed to review at least 75 cases at each centerwithin an 8-h session, including instruction time. Panels of 5

2569Eur Radiol (2021) 31:2568–2579

breast radiologists performed the review from September2016 to April 2017. Radiologists not participating in the panelcould observe the review session.

We performed a consensus-based, fully informed retro-spective radiological review. All screening and diagnostic im-ages were available to the reviewers, including ultrasound andMRI, as well as histopathological reports. The breast centerswere divided into 8 pairs; two radiologists from each centerreviewed each other’s images and made up the consensuspanel together with one independent radiologist, the first au-thor (T.H.). To qualify for the panel, we required the radiolo-gists to have at least 1 year of experience in screen reading anda reading volume of ≥ 5000 mammograms during the past 2years.

To ensure consistency across the centers in the review pro-cedures, classifications, registration, and coding of results,T.H. took part in all reviews together with a representativefrom CRN. Ahead of each session, T.H. presented the classi-fication systems and general instructions for the review. In thecase of dissent among the panel members, a majority decisionwas made. All images were reviewed locally from the picturearchiving and communication system (PACS).

Review procedure

The review procedure is described in Fig. 1. First, wereviewed the mammograms, resulting in recall and diagnosisof screen-detected breast cancer, and thereafter classifiedmammographic density using the Breast Imaging-Reportingand Data System (BI-RADS) 5th edition categories a–d [24].We identified the malignancy and classified mammographicfindings as mass, calcifications, asymmetry, distortion, or as-sociated findings using the BI-RADS lexicon. If calcificationswere present alongside another finding, the non-calcificationfinding was preferred for classification, unless calcificationswere the dominant finding. We classified the largest tumor incase of multifocality or bilateral disease. We measured thediameter (mm) of the findings on the mammogram, usingelectronic calipers. If no malignancy was visible at the timeof diagnosis, the case was classified as occult.

Thereafter, we reviewed and classified the finding, basedon its visibility on prior screening mammograms. True can-cers showed no findings at the eventual cancer site on priorscreening mammograms (Fig. 2a, b). Cancers with obviousfindings at the cancer site on priors which retrospectivelyshould have resulted in a recall, as considered by thereviewing radiologists, were defined as missed (Fig. 2c, d).Minimal sign cancers showed minor findings on prior mam-mograms, not necessarily warranting assessment (Fig. 2e, f).At review, minimal signs were classified as either actionable(recall considered possible, but not expected within a screen-ing program) or non-actionable (non-specific findings, recallnot considered possible). However, we consider all minimal

signs as one category in the main analyses. Finally, we clas-sified mammographic findings on prior mammograms formissed and minimal sign cancers.

Histopathological information extracted from the CRN da-tabase was merged with data from the review and communi-cated to the radiologists after complete classification of eachcase. Prognostic characteristics included histological type(ductal carcinoma in situ (DCIS), invasive carcinoma of nospecial type (NST), invasive lobular carcinoma (ILC), otherinvasive carcinomas) and, for invasive cancers, also histolog-ical grade, histopathological tumor diameter (mm), and lymphnode status. Predictive tumor characteristics for invasive can-cers included estrogen receptor (ER) and progesterone recep-tor (PR) status.

Statistical analyses

We performed descriptive analyses of age at diagnosis, reviewclassification categories, mammographic findings, mammo-graphic density, and histopathological characteristics. Datawere presented as percentages with 95% confidence intervals(CIs), calculated using the Clopper-Pearson mid-P interval;means ± standard deviations (SDs); and medians with inter-quartile ranges (IQRs). Chi-square tests and independent sam-ple t tests were used to test the differences between reviewclassification categories and mammographic findings or his-topathological characteristics, as well as between mammo-graphic findings and histopathological characteristics. TheBonferroni correction was used to adjust for multiple testing,and a p value < 0.001 was considered statistically significant.IBM SPSS Statistics (version 25) was used for all analyses.

Results

We reviewed and classified mammograms from 1227 womenscreened with DM, recalled due to mammographic findingsand diagnosed with breast cancer. We excluded twomammographically occult cases. The final study sample thusconsisted of 1225 cases.

Mean age at diagnosis was 62 ± 5 years, and median agewas 63 years (59, 66). We classified 46% (567/1225) of thescreen-detected cancers as true, 22% (266/1225) as missed,and 32% (392/1225) as minimal signs.

At the time of diagnosis, 59% (336/567) of true, 70% (185/266) of missed, and 65% (256/392) of minimal signs weremasses (p = 0.004 for true versus missed), while 23% (128/567) of true, 17% (46/266) of missed, and 17% (68/392) ofminimal signs were calcifications. The mean diameter ofmammographic findings at diagnosis was 18 mm ± 14 fortrue, 20 mm ± 15 for missed, and 17 mm ± 16 for minimalsigns (Table 1).

2570 Eur Radiol (2021) 31:2568–2579

Fig. 1 Review procedure

Fig. 2 a, b True screen-detected breast cancer. A 57-year-old womanpresenting with an irregular mass in the medial aspect of the left breastat screening (arrow), diagnosed with a 17-mm invasive carcinoma of nospecial type. Left craniocaudal view at diagnosis (a) and at prior screening(b). c, d Missed screen-detected breast cancer. A 67-year-old womanpresenting with a spiculated mass in the upper lateral part of the rightbreast at screening (arrow), diagnosed with an 18-mm invasive lobular

carcinoma. Right mediolateral oblique (MLO) view at diagnosis (c). Aspiculated mass at the later cancer site (arrow) also appears at priorscreening (d). e, f Minimal sign screen-detected breast cancer. A 67-year-old woman presenting with a spiculated mass in the upper outerquadrant at screening (arrow), right MLO view (e), diagnosed with a17-mm invasive lobular carcinoma. A non-specific focal asymmetry(arrow) is visible on prior screening mammograms (f)

2571Eur Radiol (2021) 31:2568–2579

At prior screening, 43% (115/266) of missed cancers and 5%(19/392) of minimal signs were masses (p < 0.001); 25% (66/266) ofmissed and 60% (236/392) of minimal signs presented asasymmetries (p < 0.001). The mean diameter of mammographicfindings on prior screening mammograms was 14 mm ± 12 formissed and 10 mm ± 9 for minimal signs (p < 0.001) (Table 1).

Eighty-two percent (247/302) of cases classified asasymmetries on prior mammograms were classified as massesat diagnosis. Ninety-nine percent (132/134) of the cases clas-sified as masses at priors were classified as masses at diagno-sis; 79% (112/142) of calcifications on priors also presented assuch at diagnosis, and 18% (26/142) of calcifications present-ed as a mass. Among cancers classified as distortions onpriors, 51% (40/78) were classified as distortions at diagnosisand 46% (36/78) as a mass (Fig. 3).

We did not observe any differences in mammographic den-sity between classification groups; the percentage of BI-

RADS a + b was 66% (372/567) for true, 67% (178/266) formissed, and 66% (260/392) for minimal signs (Table 1, Fig.4a). However, the percentages of calcifications and distortionswere statistically higher in mammograms classified with high(BI-RADS c + d) mammographic density compared with low(BI-RADS a + b), both at diagnosis (Fig. 4b) and at priorscreening (Fig. 4c).

No statistically significant differences were observed in thedistribution of histopathological type for missed, true, or min-imal signs; DCIS accounted for 17% (98/567) of the true, 12%(32/266) of the missed, and 13% (50/392) of the minimalsigns (Table 2).

Among invasive cancers, 30% (138/457) of true cancerswere histological grade 3, compared with 14% (33/230) ofmissed and 15% (51/342) of minimal signs (p < 0.001 formissed or minimal signs versus true). We observed no differ-ences in mean histopathologic tumor diameter or lymph node

Table 1 Mammographic findings at diagnosis and on prior screening mammograms stratified by review classification categories

Total True Missed Minimal signs

n = 1225 100% n = 567 46% n = 266 22% n = 392 32%

Age (years)

Mean ± SD 62 years ± 5 62 years ± 5 63 years ± 5 62 years ± 5

Median (IQR) 63 years (59, 66) 63 years (58, 66) 64 years (60, 67) 63 years (59, 66)

Mammographic density

BI-RADS a + b 810 66% (63, 69) 372 66% (62, 70) 178 67% (61, 73) 260 66% (61, 71)

BI-RADS c + d 415 34% (31, 37) 195 34% (31, 39) 88 33% (28, 39) 132 34% (29, 39)

Mammographic findings at diagnosis

Mass 777 63% (61, 66) 336 59% (55, 63) 185 70% (66, 75) 256 65% (60, 70)

Calcifications 242 20% (18, 22) 128 23% (19, 26) 46 17% (13, 22) 68 17% (14, 22)

Asymmetry 106 9% (7, 10) 58 10% (8, 13) 15 6% (3, 9) 33 8% (6, 12)

Distortion 97 8% (7, 10) 43 8% (6, 10) 20 8% (5, 11) 34 9% (6, 12)

Associated findings 3 0.2% (0, 1) 2 0.4% (0, 1) 1 0.3% (0, 1)

Mammographic diameter

Mean ± SD 18 mm ± 15 18 mm ± 14 20 mm ± 15 17 mm ± 16

Median (IQR) 14 mm (10, 21) 14 mm (10, 20) 16 mm (11, 23) 14 mm (10, 21)

Data not available 83 46 5 30

Mammographic findings prior mammograms

Mass 134 20% (17, 24) 115 43% (37, 49) 19 5% (3, 8)*

Calcifications 142 22% (19, 25) 62 23% (18, 29) 80 20% (17, 25)

Asymmetry 302 46% (42, 50) 66 25% (20, 31) 236 60% (55, 65)*

Distortion 78 12% (10, 15) 23 9% (6, 13) 55 14% (11, 18)

Associated findings 2 0.3% (0, 1) 2 0.5% (0, 2)

Mammographic diameter

Mean ± SD 12 mm ± 11 14 mm ± 12 10 mm ± 9*

Median (IQR) 9 mm (6, 14) 11 mm (7, 15) 8 mm (5, 12)

Data not available 191 20 171

Unless otherwise specified, data are the number of patients and 95% confidence intervals in parenthesis

SD standard deviation, IQR interquartile range

*p < 0.001, compared with missed

2572 Eur Radiol (2021) 31:2568–2579

status between the groups. The percentage of ER+ cases was86% (387/451) for true and 95% for both missed (215/226)and minimal signs (317/333) (p < 0.001 for missed or minimalsigns versus true). Sixty-seven percent (303/453) of true, 78%(173/223) of missed, and 77% (254/330) of minimal sign can-cers were PR+ (p = 0.004 for missed versus true and p = 0.002for minimal signs versus true) (Table 2).

We observed the highest percentage of DCIS, 60% (145/242) among cases presenting as calcifications (Table 3).Masses, asymmetries, and distortions were mainly invasivecarcinoma of NST, ranging from 71% (69/97) for distortionsto 83% (646/777) for masses. We observed 21% (20/97) forILC among distortions, compared with 7% (54/777) formasses and 2% (5/242) for calcifications. Among invasivecancers, distortions were associated with the lowest percent-age of histological grade 3 tumors (8%, 7/92) and calcifica-tions were associated with the highest (41%, 39/94). Themean mammographic diameter for invasive cancers rangedfrom 15 mm ± 8 (masses) to 36 mm ± 26 (calcifications).Mean histopathologic tumor diameter ranged from 12 mm ±11 (calcifications) to 20 mm ± 15 (distortions).

For masses, the percentage of histological grade 3 invasivecancer was higher for true than for missed and minimal sign

screen-detected breast cancer (p < 0.001 for true comparedwith both missed and minimal signs); otherwise, we observedno differences in histopathological tumor characteristics strat-ified by review classification categories and mammographicfindings (Table 4). We observed no differences for mammo-graphic findings, histologic tumor type, diameter, and gradefor minimal signs, actionable versus non-actionable tumors(Table 5 in the Appendix).

Discussion

In this informed, consensus-based review of mammogramsfrom prior screening and diagnosis of 1225 women withscreen-detected cancer, radiologists classified 46% as true,22% as missed, and 32% as minimal signs. The most frequentmammographic finding at diagnosis was a mass for all classi-fication categories; no statistically significant differences wereobserved between the classification categories regardingmammographic findings at diagnosis. At prior screening, themost frequent mammographic finding for missed cancer was amass, whereas for minimal sign cancer, it was asymmetry.The majority of asymmetries at prior screening progressed

Fig. 4 aDistribution of review classification groups based on findings onprior screening mammograms (true, missed, or minimal signs) stratifiedby the BI-RADS density score (low: BI-RADS a + b; high: BI-RADS c +d) (p = 0.88). b Distribution of mammographic findings at diagnosis

(mass, calcifications, asymmetry, or distortion) stratified by the BI-RADS density score (p < 0.001). c Distribution of mammographicfindings on prior screening mammograms stratified by the BI-RADSdensity score p < 0.001

Fig. 3 Mammographic findingson prior screening mammogramsof missed and minimal signcancers, stratified bymammographic findings onmammograms at diagnosis ofscreen-detected breast cancer

2573Eur Radiol (2021) 31:2568–2579

into masses by the time of diagnosis. Mammographic densitydid not differ between the review classification categories.True invasive cancers were more often histological grade 3and had less favorable hormonal status than missed and min-imal sign invasive cancers.

Our findings support results from other retrospective, in-formed review studies of screen-detected breast cancer. In a

study by Ikeda et al [8], findings at the later cancer site, obvi-ous or non-specific, were observed retrospectively in 67% ofscreen-detected breast cancers. Van Breest Smallenburg et al[9] found 21% of screen-detected breast cancer to be missedand 22% with non-specific minimal signs at informed review,and Broeders et al [7] identified findings on prior screeningmammograms in 53% of screen-detected cancers. However,

Table 2 Histopathological tumor characteristics stratified by review classification categories

Total True Missed Minimal signs

n = 1225 100% n = 567 n = 266 n = 392

Histopathological type

DCIS 180 15% (13, 17) 98 17% (14, 21) 32 12% (8, 17) 50 13% (10, 17)

Invasive carcinoma of NST 880 72% (69, 74) 407 72% (68, 76) 194 73% (67, 78) 279 71% (66, 76)

ILC 94 8% (6, 9) 35 6% (4, 9) 21 8% (5, 12) 38 10% (7, 13)

Other invasive carcinomas 71 6% (5, 7) 27 5% (3, 7) 19 7% (4, 11) 25 6% (4, 9)

DCIS 180 98 32 50

Tumor diameter

Mean ± SD 20 mm ± 16 20 mm ± 14 18 mm ± 14 23 mm ± 21

Median (IQR) 15 mm (10, 27) 15 mm (10, 26) 15 mm (8, 25) 17 mm (10, 33)

Data not available 23 15 2 6

Histological grade

Grade 1 23 14% (9, 21) 9 10% (5, 19) 3 10% (2, 27) 11 24% (13, 39)

Grade 2 18 11% (7, 17) 10 12% (6, 20) 3 10% (2, 27) 5 11% (4, 24)

Grade 3 121 75% (67, 81) 68 78% (68, 86) 23 79% (60, 92) 30 65% (50, 79)

Data not available 18 11 3 4

Invasive cancer 1045 469 234 342

Tumor diameter (mm)

Mean ± SD 15 mm ± 10 15 mm ± 9 16 mm ± 10 15 mm ± 10

Median (IQR) 13 mm (9, 19) 12 mm (9, 18) 13 mm (10, 20) 13 mm (9, 19)

Data not available 23 12 5 6

Histological grade

Grade 1 288 28% (25, 30) 103 23% (19, 27) 73 32% (26, 38) 112 33% (28, 38)

Grade 2 519 50% (47, 54) 216 47% (43, 52) 124 54% (47, 61) 179 52% (47, 58)

Grade 3 222 22% (19, 24) 138 30% (26, 35) 33 14% (10, 20)* 51 15% (11, 19)*

Data not available 16 12 4

Lymph node status

Negative 797 80% (77, 82) 353 78% (74, 82) 182 81% (76, 86) 262 81% (76, 85)

Data not available 47 19 10 18

Hormonal status

ER positive 919 91% (89, 93) 387 86% (82, 89) 215 95% (92, 98)* 317 95% (92, 97)*

Data not available 35 18 8 9

PR positive 730 73% (70, 75) 303 67% (62, 71) 173 78% (72, 83) 254 77% (72, 81)

Data not available 39 16 11 12

Unless otherwise specified, data are the number of patients and 95% confidence intervals in parentheses

DCIS ductal carcinoma in situ, NST no special type, ILC invasive lobular carcinoma, ER estrogen receptor, PR progesterone receptor, SD standarddeviation, IQR interquartile range

*p < 0.001, compared with true

2574 Eur Radiol (2021) 31:2568–2579

all three studies included screen-film mammography (SFM).In an informed review from BreastScreen Norway, 12% ofscreen-detected breast cancers were classified as missed and9% minimal sign actionable for screening with SFM [6] and10% missed and 9% minimal sign actionable for screeningwith DM [1]. However, the review procedures and definitionof classification groups differed between the studies, makingcomparison challenging. In experimental review studies ofinterval breast cancer exploring different study designs, thepercentage of cancers classified as missed differed largelydepending on review procedure and number of radiologists.Hofvind et al [3] found the percentage of missed intervalbreast cancer ranging from 1% (mixed, blinded individual

review) to 34% (informed, consensus-based) and 36% (in-formed individual). In a study by Ciatto et al [4], the propor-tion of missed (screening error) cancers varied from 24% in asimulated blinded review to 42% in a simulated fully informedreview. Further, studies have demonstrated that the proportionof missed cancer is affected by how close the study setting isto a normal screening setting [25, 26]. Easy understandablestandardized definitions and recommendations on classifica-tion groups and review procedure are needed to enable futurecomparison of results from reviews.

True cancers include tumors detected at an early stage,reflected by the higher percentage of DCIS among true versusmissed cancers. However, true cancers may also be fast-

Table 3 Histopathological tumor characteristics stratified by mammographic findings

Mass Calcifications Asymmetry Distortion

n = 777 n = 242 n = 106 n = 97

Histopathological type

DCIS 23 3% (2, 4)* 145 60% (53, 66) 9 9% (4, 16)* 3 3% (1, 9)*

Invasive carcinoma of NST 646 83% (80, 86)* 85 35% (29, 42) 78 74% (64, 82)* 69 71% (61, 80)*

ILC 54 7% (5, 9)** 5 2% (1, 5) 14 13% (7, 21)* 20 21% (13, 30)*

Other invasive carcinomas 54 7% (5, 9) 7 3% (1, 6) 5 5% (2, 11) 5 5% (2, 12)

Invasive cancer 754 97 97 94

Histological grade

Grade 1 211 28% (25, 32) 13 14% (8, 23) 26 28% (1, 38) 38 41% (31, 52)*

Grade 2 378 51% (47, 54) 42 45% (34, 55) 51 54% (44, 65) 47 51% (40, 62)

Grade 3 158 21% (18, 24)* 39 41% (31, 52) 17 18% (11, 27)* 7 8% (3, 15)*

Data not available 7 3 3 2

Lymph node status

Negative 583 80% (77, 83) 73 85% (76, 92) 74 80% (70, 87) 67 74% (64, 83)

Data not available 28 28 11 4 4

Hormonal status

ER positive 661 90% (88, 92) 75 85% (76, 92) 89 95% (88, 98) 92 100% (96, 100)*

Data not available 21 9 3 2

PR positive 532 73% (70, 76)* 47 54% (43, 65) 69 73% (63, 82) 80 87% (78, 93)*

Data not available 24 10 3 2

Mammographic diameter (mm)

Mean ± SD 15 mm ± 8*, ** 36 mm ± 26 17 mm ± 16* 22 mm ± 15*

Median (IQR) 13 mm (9, 18) 30 mm (17, 50) 13 mm (9, 20) 20 mm (14, 25)

Data not available 45 7 7 4

Tumor diameter (mm)

Mean ± SD 15 mm ± 8 12 mm ± 11 16 mm ± 12 20 mm ± 15*

Median (IQR) 12 mm (9, 18) 9 mm (5, 17) 14 mm (8, 17) 15 mm (12, 22)

Data not available 11 9 0 2

Unless otherwise specified, data are the number of patients and 95% confidence interval in parentheses

DCIS ductal carcinoma in situ, NST no special type, ILC invasive lobular carcinoma, ER estrogen receptor, PR progesterone receptor, SD standarddeviation, IQR interquartile range

*p < 0.001, compared with calcifications; **p < 0.001, compared with distortion

2575Eur Radiol (2021) 31:2568–2579

Table 4 Histopathological type, tumor diameter, and histological grade for mammographic findings stratified by review categories

True Missed Minimal signs

Mass n = 336 n = 185 n = 256Histopathological typeDCIS 9 3% (1, 5) 7 4% (2, 8) 7 3% (1, 6)Invasive carcinoma of NST 291 87% (83, 90) 149 81% (74, 86) 206 81% (75, 85)ILC 17 5% (3, 8) 11 6% (3, 10) 26 10% (7, 15)Other invasive carcinomas 19 6% (3, 9) 18 10% (6, 15) 17 7% (4, 10)

Tumor diameter of invasive cancerMean ± SD 14 mm ± 8 15 mm ± 9 14 mm ± 8

Median (IQR) 12 mm (9, 18) 13 mm (10, 18) 12 mm (9, 19)

Not available 5 3 3Histological grade of invasive cancerGrade 1 72 22% (18, 27) 61 35% (28, 42) 78 31% (26, 38)Grade 2 150 47% (41, 52) 96 54% (47, 62) 132 53% (47, 59)Grade 3 99 31% (26, 36) 20 11% (7, 17)* 39 16% (11, 21)*Not available 6 1 0

Asymmetry n = 58 n = 15 n = 33Histopathological typeDCIS 5 9% (3, 19) 4 12% (3, 28)Invasive carcinoma of NST 46 79% (67, 89) 10 67% (38, 88) 22 67% (48, 82)ILC 4 7% (2, 17) 15 33% (12, 62) 5 15% (5, 32)Other invasive carcinomas 3 5% (1, 14) 2 6% (1, 20)

Tumor diameter (mm) of invasive cancerMean ± SD 14 mm ± 11 20 mm ± 9 16 mm ± 14

Median (IQR) 13 mm (8, 16) 19 mm (15, 24) 14 mm (8, 16)

Not availableHistological grade of invasive cancerGrade 1 10 20% (10, 34) 2 13% (2, 41) 14 48% (29, 68)Grade 2 26 52% (37, 66) 10 67% (38, 88) 15 52% (33, 71)Grade 3 14 28% (16, 43) 3 20% (4, 48)Not available 3

Calcifications n = 128 n = 46 n = 68Histopathological typeDCIS 83 65% (56, 73) 23 50% (35, 65) 39 57% (45, 69)Invasive carcinoma of NST 39 31% (23, 39) 21 46% (31, 61) 25 37% (25, 49)ILC 3 2% (1, 7) 1 2% (0, 12) 1 2% (0, 8)Other invasive carcinomas 3 2% (1, 7) 1 2% (0, 12) 3 4% (1, 12)

Tumor diameter of invasive cancerMean ± SD 13 mm ± 12 11 mm ± 10 12 mm ± 9

Median (IQR) 10 mm (6, 15) 6 mm (4, 11) 10 mm (5, 19)

Not available 5 2 2Histological grade of invasive cancerGrade 1 6 14% (5, 27) 3 14% (3, 36) 4 14% (4, 32)Grade 2 18 41% (26, 57) 10 48% (26, 70) 14 48% (29, 68)Grade 3 20 45% (30, 61) 8 38% (18, 62) 11 38% (21, 58)Not available 1 2

Distortion n = 43 n = 20 n = 34Histopathological typeDCIS 1 2% (0, 12) 2 10% (1, 32)Invasive carcinoma of NST 29 67% (52, 81) 14 70% (46, 88) 26 77% (59, 89)ILC 11 26% (14, 41) 4 20% (6, 44) 5 15% (5, 31)Other invasive carcinomas 2 5% (1, 16) 3 9% (2, 24)

Tumor diameter (mm) of invasive cancerMean ± SD 19 mm ± 13 21 mm ±14 20 mm ± 17

Median (IQR) 16 mm (10, 22) 17 mm (12, 22) 15 mm (12, 22)

Not available 1 1Histological grade of invasive cancerGrade 1 15 37% (2, 53) 7 41% (18, 67) 16 47% (30, 65)Grade 2 22 54% (37, 69) 8 47% (23, 72) 17 50% (32, 68)Grade 3 4 10% (3, 23) 2 12% (2, 36) 1 3% (0, 15)Not available 1 1

Unless otherwise specified, data are the number of patients and 95% confidence intervals in parentheses

DCIS ductal carcinoma in situ, NST no special type, ILC invasive lobular carcinoma, ER estrogen receptor, PR progesterone receptor, SD standarddeviation, IQR interquartile range

*p < 0.001, compared with true.

2576 Eur Radiol (2021) 31:2568–2579

growing tumors, with less favorable tumor characteristics [27,28]. This is illustrated in our study by the larger percentage ofhistological grade 3 invasive cancer among true comparedwith missed and minimal sign cancers. The largest discrepan-cy in the diameter of mammographic findings versus histo-pathological tumor diameter was observed among calcifica-tions (36 mm ± 26 versus 12 mm ± 11). Invasive cancerspresenting as calcifications were also smaller and had a higherhistological grade than cancers presenting with other mammo-graphic findings. This may reflect small, aggressive invasivetumors imbedded in a larger calcified area, possibly of DCIS.However, we cannot elaborate on this hypothesis because on-ly the diameter of the invasive component is coded in thedatabase at the CRN. The larger tumor diameter and higherpercentage of tumors with low/intermediate histological gradeamong distortions may be related to the increased frequencyof ILC among distortions. ILC is generally of low/intermediate histological grade and larger at diagnosis thanother invasive carcinomas [29–31].

Amass was the most frequent finding on prior mammogramsof missed cancers, and a special emphasis onmasses at screeningmight be reasonable. Masses might be misinterpreted as benign,in particular if retrospectively visible on more than one priorscreening exam or if not having spiculated margins [18, 32].Further, themean diameter ofmammographic findings ofmissedcancer was 11 mm, which usually is regarded to be above thelimit for visual perception. This could indicate that a certainproportion of cancers was missed due to misinterpretation atscreen reading or dismissed at consensus. The high frequencyof asymmetries on priors of missed and minimal signs develop-ing into masses at the screening exam leading to diagnosis ofcancer is in line with other studies; increased awareness ofasymmetries may be useful to reduce the burden of missed can-cers at screening [32, 33]. However, asymmetries are commonand most often represent glandular tissue, in particular if visibleonly in one view. Thus, radiologists should be attentive andshould avoid an unreasonably increase in the recall rate for suchfindings. Evaluating more than one prior screening exammay bevaluable in this respect. Moreover, a recent study showed thatincreasing the recall rate mainly increased detection of low-gradeand not high-grade cancer [34]. This is consistent with our resultsdemonstrating a higher proportion of tumors of low and interme-diate histological grade among missed/minimal signs than true;an increased recall rate would probably reduce the proportion ofmissed and minimal signs.

Our study has some limitations. First, the review wasconsensus-based and all images were available—this designyields the highest percentage of missed cancers in review studies[3–5]. This limits the external generalizability of our results.Second, our study was performed at 16 breast centers with im-ages from a span of approximately 10 years, and major hetero-geneity in the combinations of PACS, workstations, and mam-mographic equipment. As a result, the image quality and

presentation differed between centers, which might have influ-enced the assessment of review classifications, mammographicfindings, and density. Third, the consensus panel included fiveradiologists: onewho participated in all the reviews and fourwhoonly participated in one session at their own and one at the pairedcenter. Although we presented and communicated the generalinstructions and classification systems to all radiologists at thestart of each review, some differences in interpretation and as-sessment between radiologists are likely to have occurred.However, we included mammograms and radiologists from allbreast centers in Norway, and our study is as far as we are awareof, the largest reported in peer-reviewed journals, which we con-sider as strength. Finally, we assessed mammographic densityfrom the screening mammograms at diagnosis. This might havebiased the association between review categories (classified fromprior screeningmammograms 2 years earlier) andmammograph-ic density, as the women’s breast density might have decreasedduring the 2 years.

The study confirmed our hypothesis that mammographicfindings and histopathologic characteristics differed betweentrue and missed screen-detected breast cancers inBreastScreen Norway. However, we did not identify differ-ences in mammographic density between review classificationcategories. One main goal of the classification of missed can-cers was quality improvement for radiologists’ performanceand the program. We would like to emphasize that the reviewand study setting differed substantially from real-life screen-ing settings. Visible findings on priors were not necessarilyindicative of a screening failure. Recalling all women withsubtle findings would increase the rate of false positive recallsand probably also the detection of small, low proliferationtumors (overdiagnosis). This is important to keep in mindduring audits and when exploring medicolegal aspects ofmammographic screening.

Acknowledgments Vestre Viken Hospital Trust funded the salary for THduring the project period. SH is the head of BreastScreen Norway.

Funding Open Access funding provided byUniversity of Oslo (incl OsloUniversity Hospital).

Compliance with ethical standards

Guarantor The scientific guarantor of this publication is Prof. SolveigHofvind.

Conflict of interest The authors of this manuscript declare no relation-ships with any companies whose products or services may be related tothe subject matter of the article.

Statistics and biometry One of the authors (K. Tsuruda) has significantstatistical expertise.

Informed consent Written informed consent was waived by the institu-tional review board. The Cancer Registry Regulation gives approval with

2577Eur Radiol (2021) 31:2568–2579

waiver of informed consent to perform surveillance, quality assurance,and studies based on data collected as a part of invitation to and partici-pation in the program

Ethical approval Institutional review board approval was obtained.Data was collected from BreastScreen Norway and thus coveredby the Cancer Registry Regulation. Data was approved by theinstitutional Data Protection Officer at the Cancer Registry andthe Heads of Department and/or research administration at thelocal breast centers.

Study subjects or cohorts overlap Some study subjects or cohorts arereported in the study of Tsuruda K, et al: Survival among women diag-nosed with breast cancer retrospectively classified as true, missed, orminimal signs through radiological review. Will be submitted toEuropean Radiology, March 2020.

Methodology• retrospective• observational• multicenter study

Appendix

Table 5 Mammographic findings and histopathological characteristics for minimal signs actionable versus minimal signs non-actionable

Minimal signs actionable Minimal signs non-actionable

n = 192 n = 200

Diagnostic mammogramsMass 140 73% (66, 79)* 116 58% (51, 65)Calcifications 24 13% (8, 18) 44 22% (17, 28)Asymmetry 13 7% (4, 11) 20 10% (6, 15)Distortion 15 8% (4, 13) 19 10% (6, 14)Associated features 1 0.5% (0, 3)Mammographic diameterMean ± SD 20 mm ± 17 18 mm ± 15Median (IQR) 15 mm (10, 22) 14 mm (9, 20)Data not available 14 16Prior mammogramsMass 13 7% (4, 11) 6 3% (1, 6)Calcifications 32 17% (12, 23) 48 24% (18, 31)Asymmetry 117 61% (54, 68) 119 60% (52, 66)Distortion 29 15% (10 21) 26 13% (9, 19)Associated features 1 0.5% (0, 3) 1 0.5% (0, 3)Histopathological typeDCIS 20 10% (7, 16) 30 15% (10, 21)Invasive carcinoma of NST 138 72% (65, 78) 141 71% (64, 77)ILC 19 10% (6, 15) 19 10% (6, 14)Other 15 8% (4, 13) 10 5% (2, 9)

Invasive cancer n = 172 n = 170Tumor diameterMean ± SD 16 mm ± 11 14 mm ± 9Median (IQR) 13 mm (9, 19) 13 mm (8, 20)Data not available 4 2Histological gradeGrade 1 54 31% (25, 39) 58 34% (27, 42)Grade 2 96 56% (48, 63) 83 49% (41, 57)Grade 3 22 13% (8, 19) 29 17% (12, 24)

Unless otherwise specified, data are the numbers of patient and 95% confidence intervals in parentheses

DCIS ductal carcinoma in situ, NST no special type, ILC invasive lobular carcinoma, SD standard deviation, IQR interquartile range

*p = 0.002, compared with minimal sign actionable

2578 Eur Radiol (2021) 31:2568–2579

Open Access This article is licensed under a Creative CommonsAttribution 4.0 International License, which permits use, sharing, adap-tation, distribution and reproduction in any medium or format, as long asyou give appropriate credit to the original author(s) and the source, pro-vide a link to the Creative Commons licence, and indicate if changes weremade. The images or other third party material in this article are includedin the article's Creative Commons licence, unless indicated otherwise in acredit line to the material. If material is not included in the article'sCreative Commons licence and your intended use is not permitted bystatutory regulation or exceeds the permitted use, you will need to obtainpermission directly from the copyright holder. To view a copy of thislicence, visit http://creativecommons.org/licenses/by/4.0/.

References

1. Hoff SR, Abrahamsen AL, Samset JH, Vigeland E, Klepp O, HofvindS (2012) Breast cancer: missed interval and screening-detected cancerat full-field digital mammography and screen-film mammography–results from a retrospective review. Radiology 264:378–386

2. Moberg K, Grundstrom H, Lundquist H, Svane G, Havervall E,Muren C (2000) Radiological review of incidence breast cancers.J Med Screen 7:177–183

3. Hofvind S, Skaane P, Vitak B et al (2005) Influence of reviewdesign on percentages of missed interval breast cancers: retrospec-tive study of interval cancers in a population-based screening pro-gram. Radiology 237:437–443

4. Ciatto S, Catarzi S, Lamberini MP et al (2007) Interval breast can-cers in screening: the effect of mammography review method onclassification. Breast 16:646–652

5. Houssami N, Hunter K (2017) The epidemiology, radiology andbiological characteristics of interval breast cancers in populationmammography screening. NPJ Breast Cancer 3:12

6. Hoff SR, Samset JH, Abrahamsen AL, Vigeland E, Klepp O, HofvindS (2011)Missed and true interval and screen-detected breast cancers ina population based screening program. Acad Radiol 18:454–460

7. Broeders MJ, Onland-Moret NC, Rijken HJ, Hendriks JH, VerbeekAL, Holland R (2003) Use of previous screening mammograms toidentify features indicating cases that would have a possible gain inprognosis following earlier detection. Eur J Cancer 39:1770–1775

8. Ikeda DM, Birdwell RL, O’Shaughnessy KF, Brenner RJ, SicklesEA (2003) Analysis of 172 subtle findings on prior normal mam-mograms in women with breast cancer detected at follow-upscreening. Radiology 226:494–503

9. van Breest SV, Setz-Pels W, Groenewoud JH et al (2012)Malpractice claims following screening mammography in theNetherlands. Int J Cancer 131:1360–1366

10. Pinto A, Acampora C, Pinto F, Kourdioukova E, Romano L,Verstraete K (2011) Learning from diagnostic errors: a good wayto improve education in radiology. Eur J Radiol 78:372–376

11. Shawky MS, Huo CW, Henderson MA, Redfern A, Britt K,Thompson EW (2019) A review of the influence of mammographicdensity on breast cancer clinical and pathological phenotype. BreastCancer Res Treat 177:251–276

12. Perry N,Broeders M de Wolf C, Tornberg S, Holland R, von KarsaL(2006) European guidelines for quality assurance in breast cancerscreening and diagnosis. Brussels, Belgium: European Communities

13. Quality manual BreastScreen Norway, Cancer Registry of Norway.Available via https://www.kreftregisteret.no/Generelt/Rapporter/Mammografiprogrammet/Kvalitet/. Accessed Dec 2019

14. NHS Breast Screening Programme (2017) Reporting, classification andmonitoringofintervalcancersandcancersfollowingpreviousassessment.Public Health England. Available via https://www.gov.uk/government/publications/breast-screening-interval-cancers. AccessedDec 2019

15. Lekanidi K, Dilks P, Suaris T, Kennett S, Purushothaman H (2017)Breast screening:What can the interval cancer review teach us?Are weperhaps being a bit too hard on ourselves? Eur J Radiol 94:13–15

16. Geertse TD, Holland R, Timmers JM et al (2015) Value of audits inbreast cancer screening quality assurance programmes. Eur Radiol25:3338–3347

17. Bansal GJ, Thomas KG (2011) Screen-detected breast cancer: doespresence of minimal signs on prior mammograms predict staging orgrading of cancer? Clin Radiol 66:605–608

18. Roberts-Klein S, Iuanow E, Slanetz PJ (2011) Avoiding pitfalls inmammographic interpretation. Can Assoc Radiol J 62:50–59

19. Bare M, Tora N, Salas D et al (2015) Mammographic and clinicalcharacteristics of different phenotypes of screen-detected and inter-val breast cancers in a nationwide screening program. BreastCancer Res Treat 154:403–415

20. Domingo L, Romero A, Blanch J et al (2013) Clinical and radiologicalfeatures of breast tumors according to history of false-positive results inmammography screening. Cancer Epidemiol 37:660–665

21. Ministry of Health and Care Services (2001) Forskrift ominnsamling og behandling av helseopplysninger i Kreftregisteret(The Cancer Registry Regulation).

22. Hofvind S, Tsuruda K,Mangerud G et al (2017) The Norwegian BreastCancerScreeningProgram,1996-2016:celebrating20yearsoforganisedmammographic screening cancer in Norway 2016 - cancer incidence,mortality,survivalandprevalenceinNorway.CancerRegistryofNorway

23. Hoff SR, Myklebust TA, Lee CI, Hofvind S (2019) Influence ofmammography volume on radiologists’ performance: results fromBreastScreen Norway. Radiology 292:289–296

24. Sickles E, D’Orsi CJ, Bassett LW et al (2013) ACRBI-RADS®mam-mography. In: ACR BI-RADS® atlas, Breast Imaging Reporting andData System. American College of Radiology, Reston

25. Moberg K, Grundstrom H, Tornberg S et al (1999) Two models forradiological reviewing of interval cancers. J Med Screen 6:35–39

26. Gordon PB, Borugian MJ, Warren Burhenne LJ (2007) A truescreening environment for review of interval breast cancers: pilotstudy to reduce bias. Radiology 245:411–415

27. Domingo L, Salas D, Zubizarreta R et al (2014) Tumor phenotypeand breast density in distinct categories of interval cancer: results ofpopulation-based mammography screening in Spain. Breast CancerRes 16:R3

28. Hofvind S, Sagstad S, Sebuodegard S, Chen Y, Roman M, Lee CI(2018) Interval breast cancer rates and histopathologic tumor char-acteristics after false-positive findings at mammography in apopulation-based screening program. Radiology 287:58–67

29. Chamming’s F, Kao E, Aldis A et al (2017) Imaging features andconspicuity of invasive lobular carcinomas on digital breasttomosynthesis. Br J Radiol 90:20170128

30. Chen Z, Yang J, Li S et al (2017) Invasive lobular carcinoma of thebreast: a special histological type compared with invasive ductalcarcinoma. PLoS One 12:e0182397

31. Li CI, Uribe DJ, Daling JR (2005) Clinical characteristics of differ-ent histologic types of breast cancer. Br J Cancer 93:1046–1052

32. Wadhwa A, Sullivan JR, Gonyo MB (2016) Missed breast cancer:what can we learn? Curr Probl Diagn Radiol 45:402–419

33. Chesebro AL, Winkler NS, Birdwell RL, Giess CS (2016)Developing asymmetries at mammography: a multimodality ap-proach to assessment and management. Radiographics 36:322–334

34. Blanks RG, Given-Wilson RM, Cohen SL, Patnick J, Alison RJ,WallisMG (2019) An analysis of 11.3 million screening tests examining theassociation between recall and cancer detection rates in the EnglishNHSbreast cancer screening programme. Eur Radiol 29:3812–3819

Publisher’s note Springer Nature remains neutral with regard to jurisdic-tional claims in published maps and institutional affiliations.

2579Eur Radiol (2021) 31:2568–2579